Compiler Internals

Hello,

What would be considered a good choice for the intermediate representation
in a scheme compiler these days? So far I've found CPS, ANF, and SSA
with them being translatable between each other, with exceptions. I
understand SSA and CPS the best and am just looking into ANF.

There were a great crop of theises and papers around the mid 90's about
this stuff especially control/data flow determination with CPS, but I
can't seem to find much newer stuff (including at readscheme.org).

Anyone have any good pointers to some recent papers about this topic?

Thank you.

-pete

On 28 Aug 2008 18:39:18 GMT, Peter Keller <psilord@merlin.cs.wisc.edu>
wrote:

>Hello,
>
>What would be considered a good choice for the intermediate representation
>in a scheme compiler these days? So far I've found CPS, ANF, and SSA
>with them being translatable between each other, with exceptions. I
>understand SSA and CPS the best and am just looking into ANF.

CPS tends to be the preferred representation. SSA can get quite hairy
with general continuations. I've read a lot about ANF but I've never
actually seen a compiler use it.


>There were a great crop of theises and papers around the mid 90's about
>this stuff especially control/data flow determination with CPS, but I
>can't seem to find much newer stuff (including at readscheme.org).
>
>Anyone have any good pointers to some recent papers about this topic?

Sorry, I don't have anything recent to give you. You might ask in
comp.compilers ... they're usually up on the latest research.


George

Hi Peter

> There were a great crop of theises and papers around the mid 90's about
> this stuff especially control/data flow determination with CPS, but I
> can't seem to find much newer stuff (including at readscheme.org).
>
> Anyone have any good pointers to some recent papers about this topic?

One such pointer is:

Andrew Kennedy
"Compiling with continuations, continued"
ICFP'07

available here:

http://research.microsoft.com/~akenn...sContinued.pdf


Enjoy,


//Jan

Hello,

Jan Midtgaard <hestetest@hotmail.com> wrote:
> One such pointer is:
>
> Andrew Kennedy
> "Compiling with continuations, continued"
> ICFP'07
>
> available here:
>
> http://research.microsoft.com/~akenn...sContinued.pdf

Thanks! That's the type of thing for which I'm looking.

-pete

On 31 ago, 17:30, Jan Midtgaard <hestet...@hotmail.com> wrote:
> Hi Peter
>
> > There were a great crop of theises and papers around the mid 90's about
> > this stuff especially control/data flow determination with CPS, but I
> > can't seem to find much newer stuff (including at readscheme.org).
>
> > Anyone have any good pointers to some recent papers about this topic?
>
> One such pointer is:
>
> * *Andrew Kennedy
> * *"Compiling with continuations, continued"
> * *ICFP'07
>
> available here:
>
> *http://research.microsoft.com/~akenn...ontinuationsCo....
>
> Enjoy,
>
> //Jan

Nice to see CPS still going strong. The author explicitely adresses
previous criticisms in well-known papers.

Peter Keller wrote:
> Hello,
>
> Jan Midtgaard <hestetest@hotmail.com> wrote:
>> One such pointer is:
>>
>> Andrew Kennedy
>> "Compiling with continuations, continued"
>> ICFP'07
>>
>> available here:
>>
>> http://research.microsoft.com/~akenn...sContinued.pdf


Title:
Amortizing the cost of commuting conversions
when beta-reducing monadic normal forms and A-normal forms
Author:
Olivier Danvy, University of Aarhus
Abstract:
In "Compiling with Continuations, Continued" (ICFP 2007), Andrew
Kennedy points out that commuting conversions increase the complexity
of simplifying intermediate-language terms, and states that "it is
far from clear how to amortize the cost of commuting conversions to
obtain a linear number of reductions for A-normal forms", in contrast
to CPS. We show how to achieve this amortization, as a corollary of
a syntactic bijection between CPS terms and terms in A-normal form.

See the IFL Proceedings from Freiburg:

http://proglang.informatik.uni-freib...roceedings.pdf


--
Jens Axel Søgaard

Jens Axel Soegaard <findrealaddresswithgoogle@soegaard.net> wrote:
> See the IFL Proceedings from Freiburg:
>
> http://proglang.informatik.uni-freib...roceedings.pdf

I didn't even know about the IFL Proceedings. That is very useful to me.

Thanks!

-pete